I have been working through the ASF's "Recipes" for the creation of a DSM from Sentinel 1 (A and B) imagery.
I am able to produce a DSM, and am relatively happy with the results in terms of the surface produced - it appears to follow the trends of my known datasets (hills where there should be hills and valleys where there are valleys).
I am however seeing discrepancies between my derived surface and a known point (derived from LiDAR). I generated a DSM using imagery acquired in the same month (a few days either side to improve the Baseline and remove temporal discrepancies) as when my LiDAR dataset was acquired.
I am not going to dispute that the LiDAR Elevation is going to give me a better Z-Value than a Satellite derived DSM. I am however looking to use Sentinel in areas where there is not a decent surface or topographic dataset I can use.
I would like to understand why there is a difference (consistently around 120m) between my DSM and LiDAR surfaces.
How can I improve the vertical accuracy of my DSM? Is the issue a vertical datum issue or perhaps a setting / attribute I should be looking for in the acquisition of the imagery?